Click through rate estimation in varying display situations

Abstract

The presently disclosed subject matter includes at least a method, system and a program storage device of estimating click through rate (CTR) of a pair of source and recommendation, the source comprising a plurality of slots, each slot configured to present a served recommendation. The CTR estimation is performed by first determining an estimated calibration CTR coefficient for each slot and then using this information for calculating an estimated CTR of a given slot when served with a given recommendation in a given source.

The invention claimed is: 
     
       1. A computer implemented method of estimating click through rate (CTR) of a recommendation, the recommendation being presented in at least one source; the source comprising a plurality of slots, each slot configured to present a served recommendation; wherein the computer comprises at least one processor and is operatively connected to a data-repository, the method comprising, with the help of the at least one processor:
 performing a calibration process comprising: 
 serving recommendations randomly to slots presented in page views of a calibration group; for each slot: 
 monitoring the served recommendations on the slot and maintaining a first counter for counting the number of recommendations which are served in a slot; 
 monitoring clicks on the slot and maintaining a second counter for counting the number of clicks on recommendations served in the slot; 
 determining an estimated calibration CTR coefficient for the slot based on a ratio between the first counter and the second counter; 
 performing data collection process, comprising:
 for a pair of a given source and a given recommendation, serving the given recommendation in different slots in the given source in multiple page views; 
 for a given recommendation, serving the given recommendation in multiple page views in different slots of one or more sources; 
 
 maintaining in the data repository a respective third counter for counting the number of clicks on the given recommendation when served in any one of the one or more sources; 
 maintaining in the data repository a respective fourth counter such that each time the given recommendation is served in any given slot of any one of the one or more sources, the CTR coefficient of the given slot is added to the fourth counter; 
 wherein serving the recommendation in one or more sources, each source having one or more slots, does not increase the number of the third counters and the number of the fourth counters assigned to each recommendation; and 
 determining an estimated CTR of the given recommendation when served in a certain slot based on the estimated calibration CTR coefficient of the certain slot and the ratio between the fourth counter and the third counter. 
 
     
     
       2. The method according to  claim 1  wherein the calibration group comprises a part of all page views of all sources. 
     
     
       3. The method according to  claim 1  wherein the counters are updated in real-time. 
     
     
       4. The method according to  claim 1  wherein the size of the calibration group is a varying size. 
     
     
       5. The method according to  claim 1  wherein the estimated calibration CTR coefficient is determined while taking into consideration additional information which influences the number of clicks on the slot. 
     
     
       6. The method according to  claim 1  further comprising:
 defining, a recent time period and a previous time period; 
 maintaining a recent first counter and a respective recent second counter for the recent time period and determining a recent estimated calibration CTR coefficient; 
 maintaining a previous first counter and a previous recent second counter for the recent time period and determining a previous estimated calibration CTR coefficient; 
 performing a statistical test for determining whether the recent estimated calibration CTR coefficient and the previous estimated calibration CTR coefficient are the same or not; and 
 in case it is determined that the recent estimated calibration CTR coefficient and the previous estimated calibration CTR coefficient are not the same, determining an estimated time of change in the display layout of the slot. 
 
     
     
       7. The method according to  claim 6  further comprising using the recent estimated calibration CTR coefficient for estimating CTR based on information obtained during the recent time period. 
     
     
       8. The method according to  claim 6  further comprising using the previous estimated calibration CTR coefficient for estimating CTR based on information obtained during the previous time period. 
     
     
       9. The method according to  claim 6  wherein time is divided into time windows of fixed size, the recent time period being a recent time window and the previous time period being a previous time period. 
     
     
       10. The method according to  claim 6 , wherein time is divided into time windows of fixed size, the method further comprising:
 defining the previous time period as a time period from a time window of a last detected change time stamp to a given time window and defining the recent time period as a time period starting after the given time window to a recent time window. 
 
     
     
       11. The method according to  claim 10  further comprising:
 dividing the time windows multiple times, wherein in each division the size of the previous time period and the size of the recent time period is different; 
 performing the statistical test for each division until a predefined number of divisions is reached and/or until the statistical test shows a sufficient difference between the previous time period and the recent time period. 
 
     
     
       12. The method according to  claim 1  further comprising selecting page views or users for the calibration group. 
     
     
       13. The computerized method according to  claim 1  wherein the one or more source include only one source. 
     
     
       14. A non-transitory program storage device readable by machine, tangibly embodying a program of instructions executable by the machine to perform operations of a method of estimating click through rate (CTR) of a recommendation, the recommendation being presented in at least one source; the source comprising a plurality of slots, each slot configured to present a served recommendation, the storage device comprising:
 instructions for performing a calibration process comprising: 
 instructions for serving recommendations randomly to slots presented in page views of a calibration group; for each slot: 
 instructions for monitoring the served recommendations on the slot and maintaining a first counter for counting the number of recommendations which are served in a slot; 
 instructions for monitoring clicks on the slot and maintaining a second counter for counting the number of clicks on recommendations served in the slot; 
 instructions for determining an estimated calibration CTR coefficient for the slot based on a ratio between the first counter and the second counter; 
 instructions for performing data collection process, comprising: 
 instructions that, for a pair of a given source and a given recommendation, serve the given recommendation in different slots in the given source in multiple page views; 
 instructions that, for a given recommendation, serve the given recommendation in multiple page views in different slots of one or more sources; 
 instructions for maintaining in a data repository a respective third counter for counting the number of clicks on the given recommendation when served in any one of the one or more sources; 
 instructions for maintaining in the data repository a respective fourth counter such that each time the given recommendation is served in any given slot of any one of the one or more sources, the CTR coefficient of the given slot is added to the fourth counter; 
 wherein serving the recommendation in one or more sources, each source having one or more slots, does not increase the number of the third counters and the number of the fourth counters assigned to each recommendation; and 
 instructions for determining an estimated CTR for of the given recommendation when served in a certain slot based on the estimated calibration CTR coefficient of the certain slot and the ratio between the fourth counter and the third counter. 
 
     
     
       15. The program storage device according to  claim 14  further comprising:
 instructions for defining, a recent time period and a previous time period; 
 instructions for maintaining a recent first counter and a respective recent second counter for the recent time period and determining a recent estimated calibration CTR coefficient; 
 instructions for maintaining a previous first counter and a previous recent second counter for the recent time period and determining a previous estimated calibration CTR coefficient; 
 instructions for performing a statistical test for determining whether the recent estimated calibration CTR coefficient and the previous estimated calibration CTR coefficient are the same or not; 
 instructions that, in case it is determined that the recent estimated calibration CTR coefficient and the previous estimated calibration CTR coefficient are not the same, determine an estimated time of change in the display layout of the slot; and 
 instructions for using the recent estimated calibration CTR coefficient for estimating CTR based on information obtained during the recent time period. 
 
     
     
       16. The program storage device according to  claim 15  further comprising:
 instructions for using the previous estimated calibration CTR coefficient for estimating CTR based on information obtained during the previous time period. 
 
     
     
       17. The program storage device according to  claim 15  further comprising:
 dividing time into time windows of fixed size the recent time period being a recent time window, and the previous time period being a previous time period. 
 
     
     
       18. The program storage device according to  claim 15  the operations of the method further comprising:
 instructions for dividing time into time windows of fixed size; 
 instructions for defining the previous time period as a time period from a time window of a last detected change time stamp to a given time window and defining the recent time period as a time period starting after the given time window to a recent time window. 
 
     
     
       19. The program storage device according to  claim 18  further comprising:
 instructions for dividing the time windows multiple times, wherein in each division the size of the previous time period and the size of the recent time period is different; 
 instructions for performing a statistical test for each division until a predefined number of divisions is reached and/or until the statistical test shows a sufficient difference between the previous time period and the recent time period. 
 
     
     
       20. A computerized system, configured to estimate click through rate (CTR) of a recommendation, the recommendation is presented in at least one source, the source comprising a plurality of slots, each slot configured to present a served recommendation; the system comprising at least one processor operatively connected to a computer data-repository, the at least one processor being configured to:
 perform a calibration process comprising: 
 serve recommendations randomly to slots presented in page views of a calibration group; for each slot: 
 monitor the served recommendations on the slot and maintaining a first counter for counting the number of recommendations which are served in a slot; 
 monitor clicks on the slot and maintaining a second counter for counting the number of clicks on recommendations served in the slot; 
 determine an estimated calibration CTR coefficient for the slot based on a ratio between the first counter and the second counter; 
 perform data collection process, comprising: 
 for a pair of a given source and a given recommendation, serve the given recommendation in different slots in the given source in multiple page views; 
 for a given recommendation, serving the given recommendation in multiple page views in different slots of one or more sources; 
 maintain in the data repository a respective third counter for counting the number of clicks on the given recommendation when served in any one of the one or more sources; 
 maintain in the data repository a respective fourth counter such that each time the given recommendation is served in any given slot of any one of the one or more sources, the CTR coefficient of the given slot is added to the fourth counter; 
 wherein serving the recommendation in one or more sources, each source having one or more slots, does not increase the number of the third counters and the number of the fourth counters assigned to each recommendation; and 
 determine an estimated CTR of the given recommendation when served in a certain slot based on the estimated calibration CTR coefficient of the certain slot and the ratio between the fourth counter and the third counter. 
 
     
     
       21. The computerized system according to  claim 20  wherein the at least one processor is further configured to:
 define, a recent time period and a previous time period; 
 maintain a recent first counter and a respective recent second counter for the recent time period and determining a recent estimated calibration CTR coefficient; 
 maintain a previous first counter and a previous recent second counter for the recent time period and determining a previous estimated calibration CTR coefficient; 
 perform a statistical test for determining whether the recent estimated calibration CTR coefficient and the previous estimated calibration CTR coefficient are the same or not; 
 in case it is determined that the recent estimated calibration CTR coefficient and the previous estimated calibration CTR coefficient are not the same, determine an estimated time of change in the display layout of the slot; and 
 utilize the recent estimated calibration CTR coefficient for estimating CTR based on information obtained during the recent time period. 
 
     
     
       22. The computerized system according to  claim 21  the at least one processor is further configured to:
 utilize the previous estimated calibration CTR coefficient for estimating CTR based on information obtained during the previous time period. 
 
     
     
       23. The computerized system according to  claim 21  the at least one processor is further configured to:
 divide time into time windows of fixed size, the recent time period being a recent time window and the previous time period being a previous time period. 
 
     
     
       24. The computerized system according to  claim 21  further configured to:
 divide time into time windows of fixed size; define the previous time period as a time period from a time window of a last detected change time stamp to a given time window and defining the recent time period as a time period starting after the given time window to a recent time window. 
 
     
     
       25. The computerized system according to  claim 24  the at least one processor is further configured to:
 divide the time windows multiple times, wherein in each division the size of the previous time period and the size of the recent time period is different; and 
 perform the statistical test for each division until a predefined number of divisions is reached and/or until the statistical test shows a sufficient difference between the previous time period and the recent time period. 
 
     
     
       26. The computerized system according to  claim 20  wherein the computer memory comprises a calibration module configured with instructions for performing the calibration process and an estimation module configured with instruction for performing the data tracking process and determining of an estimated CTR.